Random forest (RF) is widely regarded as one of the most prevalent machine learning algorithms. To achieve higher precision, the structure of decision trees that serve as base learners in RF has become increasingly complex. Consequently, the complication of RF's rules has led to a degradation in interpretability. To address this issue, the fuzzy bicausal random forest (FBRF) model is proposed. A novel causal evaluation metric, named fuzzy bi-directional causality recognition or fuzzy bicausality recognition (FBR), has been proposed to measure a more comprehensive causal relationship combining with fuzzy theory. FBR constructs a causality estimation model from both sufficient and necessary perspectives, and introduces fuzzy theory to handle the uncertainty in the process of feature discretization. Building on this foundation, decision trees and RFs are established with FBR to form the FBRF while conducting complexity analysis simultaneously. In terms of experiments, results from twenty public datasets of different sizes show that FBRF simplifies the model scale while maintaining comparable precision compared to other RF variants. Therefore, it is demonstrated that FBRF reaches a trade-off between superior interpretability and high performance.